Due to its vast range of applications, vehicle license plate recognition (VLPR) is a critical research topic in the field of intelligent transportation systems (ITS). VLPR can be utilized to help humans to automatically detect plate number without human supervision. VLPR has been used extensively in highway and bridge charge, port, and airport gate monitoring. License plates usually possess distinct features in terms of color, size, and shape in each country. Different techniques have been developed for license plate detection and recognitions. All of that motivates our work in implementing an efficient VLPR system adapted to Libyan license plates in order to profit of various applications of VLPR systems in Libya. An end-to-end VLPR system based on a neural network with histogram of oriented gradient (HOG) features is presented in this research. The suggested LPR system is divided into the following stages: localization, segmentation, and recognition. The suggested system’s sensitivity was evaluated on a database of fixed images and video sequences. For analyzing the performance of the suggested system, the optimal set of parameters that provides the best results has been evaluated. The sensitivity of the localization and recognition stages has been tested based on the adjustment of the primary parameters of the proposed system to identify the optimum combination of parameters that offers the best results. The proposed localization method is compared to the zero-crossing method, which is another localization method. The comparison findings demonstrate that the proposed method provides better overall performance.